遇见数据集

Galaxy Zoo 2: Images from Original Sample

收藏
Zenodo2020-07-30 更新2026-05-25 收录
数据链接:
官方服务:

资源简介:

The Galaxy Zoo team regularly receives requests for subject images for various versions of Galaxy Zoo, in order to facilitate other investigations, e.g. machine learning projects. This repository is an updated attempt to provide those in a way that is useful to the wider community. The images here are meant to be used with the data tables available at data.galaxyzoo.org. They are the "original" sample of subject images in Galaxy Zoo 2 (Willett et al. 2013, MNRAS, 435, 2835, DOI: 10.1093/mnras/stt1458) as identified in Table 1 of Willett et al. and also in Hart et al. (2016, MNRAS, 461, 3663, DOI: 10.1093/mnras/stw1588). The original GZ2 subjects also gave the option to view an inverted version of the subject image; these inverted images are not provided but are easily reproducible from the included subject images. <strong>If you use this dataset, please cite</strong> Willett et al. (2013) as the general data release and <em>also</em> cite the DOI for this dataset; if you use the updated debiased tables from Hart et al. (2016) please cite that as well. There are 243,434 images in total. This is off by about 0.08% from the total count in the tables - it's not clear what the cause of the discrepancy is, but we don't think the missing images have any particular sampling bias, so this sample should be useful for research. The images are available in a single zip file (<strong>images_gz2.zip</strong>). The most recent and reliable source for morphology measurements is "GZ2 - Table 1 - Normal-depth sample with new debiasing method – CSV" (from Hart et al. 2016), which is available at data.galaxyzoo.org. To cross-reference the images with Table 1, this sample includes another CSV table (<strong>gz2_filename_mapping.csv</strong>) which contains three columns and 355,990 rows. The columns are: <strong>objid</strong>: the Data Release 7 (DR7) object ID for each galaxy. This should match the first column in Table 1. <strong>sample</strong>: string indicating the subsampling of the galaxy. <strong>asset_id</strong>: an integer that corresponds to the filename of the image in the zipped file linked above. As an example row: 587722981742084144,original,16 The galaxy is 587722981741363294, which is in Table 1 and was identified by GZ2 volunteers as a barred spiral galaxy with a mild bulge and two tightly-wound arms (morphology='Sc2t'). It is in the original GZ2 sample, and can be found in the zipped file as 16.jpg. The overlap between the set of images, the attached table, and Table 1 is not 100%; there are a few rows in the tables that don't have a corresponding image. Again, it's not clear what the exact reason is for this, but we suggest just dropping any missing rows/images from your analysis unless you have a need for analyzing specific subjects. If you do need a 100% complete sample, you can obtain the missing images directly from SDSS. Based on spot checks the mappings between asset ID and DR7 object ID appear correct, but we strongly suggest that you pick some random images and verify on your own that the image seems to match the label/classifications that are listed in Table 1. If you have any issues using this dataset, please contact the Galaxy Zoo team, in particular Brooke Simmons (b.simmons@lancaster.ac.uk). Should Dr Simmons be unavailable, try contacting Karen Masters or Chris Lintott. - the GZ team, 5 Dec 2019<br>

Galaxy Zoo(星系动物园)团队经常收到来自Galaxy Zoo各版本的待测图像请求,以助力各类后续研究(例如机器学习项目)。本仓库旨在以更便于广泛社区使用的方式,提供这些图像资源,为此进行了更新优化。 此处提供的图像需配合data.galaxyzoo.org上的数据集表格使用。这些图像属于Galaxy Zoo 2(GZ2)的原始待测图像样本,对应Willett等人2013年发表于《MNRAS》第435卷第2835页的研究(DOI: 10.1093/mnras/stt1458),同时也符合Willett等人的表1以及Hart等人2016年发表于《MNRAS》第461卷第3663页的研究(DOI: 10.1093/mnras/stw1588)中的界定标准。 原始GZ2待测图像还提供了反色显示版本的可选功能,但本仓库未提供该反色图像,不过可通过附带的待测图像轻松生成。 <strong>若使用本数据集,请务必引用</strong>Willett等人2013年的研究作为通用数据发布依据,<em>同时</em>引用本数据集的DOI;若使用Hart等人2016年更新的去偏表格,请一并引用该研究。 本数据集共包含243,434张图像,与表格中的总计数存在约0.08%的偏差。目前尚未明确该差异的成因,但我们认为缺失的图像不存在特定的抽样偏差,因此该样本仍可用于相关研究。 图像打包于单个zip压缩文件(<strong>images_gz2.zip</strong>)中。目前最权威的形态测量数据源为"GZ2 - Table 1 - Normal-depth sample with new debiasing method – CSV"(源自Hart等人2016年的研究),可从data.galaxyzoo.org获取。 为实现图像与表1的交叉对照,本数据集附带了另一张CSV表格(<strong>gz2_filename_mapping.csv</strong>),该表格包含3列,共355,990行。各列说明如下:<strong>objid</strong>:每个星系的第7次数据发布(Data Release 7,DR7)天体ID,应与表1的第一列匹配。<strong>sample</strong>:字符串类型,指示该星系的子采样分组。<strong>asset_id</strong>:整数类型,与上述压缩包内图像文件的文件名对应。 示例行如下:587722981742084144,original,16。该条目对应的星系为587722981741363294,该星系出现在表1中,且被GZ2志愿者归类为具有平缓核球和两条紧密缠绕旋臂的棒旋星系(形态分类为'Sc2t')。该星系属于原始GZ2样本,可在压缩包中以16.jpg文件名找到。 图像集合、附带表格与表1之间并非完全一一对应;部分表格行未找到对应的待测图像。目前尚未明确该现象的确切成因,我们建议在分析中直接丢弃缺失的行或图像,除非您有特定需求需要分析特定待测样本。若您需要100%完整的样本,可直接从斯隆数字巡天(SDSS)获取缺失的图像。 经抽样检查,asset_id与DR7天体ID的映射关系均正确,但我们强烈建议您随机抽取部分图像,自行验证图像是否与表1中列出的标签/分类结果匹配。 若您在使用本数据集时遇到任何问题,请联系Galaxy Zoo(星系动物园)团队,特别是Brooke Simmons(邮箱:b.simmons@lancaster.ac.uk)。若联系不到Simmons博士,可尝试联系Karen Masters或Chris Lintott。——GZ团队,2019年12月5日<br>

提供机构:
Zenodo
创建时间:
2019-12-19
二维码
社区交流群
二维码
科研交流群
商业服务